diplomacy, precrisis communication, and war · diplomatic communication generally increases the...

Diplomacy, Precrisis Communication, and War

Shuhei Kurizaki

Texas A&M University

October 17, 2011

1 Book Project

2 Informational Origins of Diplomacy and War

3 The Model

4 Equilibria

5 Credibility of Diplomatic Communication

6 Diplomacy and the Risk of War

7 Conclusion

8 APPENDIX

Project on Diplomacy

Diplomacy is a primary form of “Politics Among Nations”


Diplomacy is one of the oldest political institutions designed topreserve order in the international society (Hedley Bull 1977).


Diplomacy has evolved as a set of mechanisms for conflictresolution as an alternative to military might(Hans Morgenthau 1973).


IR scholarship has sidestepped the study of diplomacy


Ultimate Goal

Bring diplomacy back into the IR literature, and establish the study ofdiplomacy as a full-fledged social scientific literature.

Book Project

When Diplomacy Works

This book offers the first comprehensive theory of diplomacy and itsrole in international disputes

Emphasizes theoretical foundations, but offers historical andstatistical materials

Diplomacy is an institutional response to bargaining failure

Diplomacy is a new addition to the “Correlates of War”

Book Project

Three classes of diplomatic mechanisms

Cheap-talk communication in pre-crisis diplomacy

Communication

Pre-crisis diplomacy

Conflict of interests Diplomatic demands

Book Project



Diplomatic negotiation in international disputes

Negotiation

Diplomatic negotiation

Military threats Diplomatic demands

Book Project



Diplomatic negotiation in international disputes

Diplomatic manipulation in military crises

Manipulation

Crisis diplomacy

Military threats Outbreak of war

Diplomacy, Precrisis Communication, and War

Shuhei Kurizaki

Texas A&M University

October 17, 2011

1 Book Project


3 The Model

4 Equilibria



7 Conclusion

8 APPENDIX

Informational Origins of Diplomacy and War

Informational Origin of War

War occurs when political or military information is privateinformation

States have incentives to misrepresent their private information

Incomplete information gives rise to the risk-return trade-off

A positive probability of war


One “solution” to incomplete information is to let them talkbefore crisis bargaining

Diplomacy offers institutional solutions to the informationalcause of war


The Byzantine Empire

Aftermath of the fall of the Roman empire and military deficit

Turned to various manipulative tactics and information

Bureau of Barbarians, the first intelligence agency


Italian city-states during the Renaissance

The system of resident, permanent embassies

To stabilize and facilitate communication among governments

The fear of invasions by the Ottoman Empire and others

To overcome mutual mistrust and enhance collective security effort


Cardinal Richellieu

“Continuous” diplomacy

Forestall potential frictions or tensions

Professional diplomats & first ministry of foreign affairs


Scholars of IR also have envisioned this informational“solution” as a key function of diplomacy

Hans Morgenthau

A main task of diplomacy is to determine one’s foreign policyobjectives in light of the military capability (Prudence). . . and to assess objectives and motivations of other nationsand their military capability.

Conventional Wisdom

Rationalist Approach

Informational problem in pre-crisis communication

Pessimistic conclusions

Conventional Wisdom

Cheap-Talk Diplomacy (Fearon 1995)

This “solution” is impossible

Diplomacy is ineffective: No info. credibly conveyedDiplomacy is irrelevant: No effect on crisis behavior

Common wisdom for a long time

Conventional Wisdom

Recent formal models

Institutional arrangements permit:

Only limited communication (Sartori, Ramsay)

Full communication only under grim-trigger (Guisinger and Smith)

When informative, always increases the risk of war

Only exception is Trager (2010)

Two Puzzles

Credibility of diplomatic communication

Puzzle: Architects and theorists all envisioned diplomacy as aninformation-revelation device. Are they wrong about that?

Two Puzzles

Diplomacy and the risk of war

Puzzle: Diplomacy has been used for hundreds of years, and many stateshave invested significant resources to keeping diplomatic institutions. Ifdiplomacy is dangerous, why keep it?

My Contributions

In this paper . . .

Use the same analytical focus of the previous models

Generalize the previous models

Refine the analysis

Find new, improved answers.

Provide better understanding on the role of diplomacy.

1 Book Project


3 The Model

4 Equilibria



7 Conclusion

8 APPENDIX

The Model, overview

State 1 and State 2 are in a dispute over some good of size vi > 0,for i = 1, 2

Resist

1

Surrender

2pv1 – c1(1 – p)v2 – c2

(v1, 0)(0, v2)Status quo

Attack

The Model, overview

Before the crisis, State 2 sends a cheap-talk messagem ∈ {resist, surrender}

Resist

1

Surrender

22

m {r, s}

pv1 – c1(1 – p)v2 – c2

(v1, 0) (0, v2)Status quo

Attack

The Model, overview

Stage game is indefinitely repeated with δ ∈ (0, 1).

Resist

1

Surrender

22

m {r, s}

pv1 – c1(1 – p)v2 – c2


Attack

1 Book Project


3 The Model

4 Equilibria



7 Conclusion

8 APPENDIX

Equilibrium: Trigger Strategy

Normal Phase

S2 maintains normal diplomatic relations with S1.

Punishment Phase

S1 suspends diplomatic relations with S2, so ignores m

Lasts for N periods

N = 0 (Fearon 1995)N = 2 (Sartori 2002)N = ∞ (Guisinger & Smith 2002)

Diplomatic normalization occurs if S2 plays the equilibrium strategyduring diplomatic punishment.

Equilibrium: Two Cases

Babbling Equilibria

Diplomacy conveys no meaningful information

Crisis behavior is not influenced

Risk of war remains the same as in the game without diplomacy

Basis for coercive diplomacy

Informative Equilibria

Diplomacy conveys no meaningful information

Diplomacy can increase the risk of war

Diplomacy can decrease the risk of war

Basis for the “peace by peaceful means”

Babbling Equilibria

S2 randomizes its diplomatic messages

Babbling Equilibria

No information is conveyed

Crisis behavior not influencedSame ex ante risk of war as in the game without diplomacy

Babbling Equilibria

Diplomatic mechanisms “turned off”

Equivalent to one-shot game ⇔ δ = 0=⇒ Conventional wisdom holds only in the absence of diplomaticinstitutions


Information conveyed

Deception equilibrium: Commitments are partially revealedSincere equilibrium: Commitments are fully revealed


Diplomacy affects crisis behavior


Threats to break off diplomatic relations induce truth-telling

Babbling equilibria constitute the Nash (reversion) threat


Describe how diplomatic communication and representation cometogether to form diplomatic institutions.

Use the informative equilibria to ask

Question about the informativeness of diplomacy and itsconditionsQuestion about its effect on the risk of war

1 Book Project


3 The Model

4 Equilibria



7 Conclusion

8 APPENDIX

Credibility of Diplomatic Communication

Credibility of Diplomacy and Equilibrium Types

⎧⎪⎨⎪⎩

Diplomacy is not credible in the babbling equilibria

Diplomacy is partially credible in the deception equilibrium

Diplomacy is fully credible in the sincere equilibrium

Credibility conditions � Equilibrium condition


Equilibrium Condition: Common Discount Factor

Let δ = c2G(κ(r))(1−p(1−G(κ(r))))(w2(H)−w2(D1)) .

⎧⎪⎨⎪⎩

Babbling equil exist if δ = 0

Deception equil can exist if 0 < δ < δ

Sincere equil can exist if δ ≥ δ


Minimum discount factor δ to sustain the sincere equilibrium.

0 2 4 6 8 10

0.0

0.2

0.4

0.6

0.8

1.0

Suspension duration, N

Min

imum

del

ta, δ

p = 0.1

p = 0.3p = 0.5

p = 0.7


Equilibrium Condition: Punishment Duration

Let N = ln(δ−ζ(1−δ))ln(δ) − 1.⎧⎪⎨⎪⎩

Only Babbling equil exist if N = 0

Deception equil can exist if 0 < N < N

Sincere equil can exist if N ≥ N

where ζ = G(κ∗)c2

(1−p(1−G(κ∗)))(π(C)−π(D))


Minimum punishment length to sustain the Sincere Equilibrium

0.0 0.2 0.4 0.6 0.8

02

46

810

1214

Capability ratio, p, and Discount factor, δ

Min

imum

sus

pens

ion,

N

δ, c=0.2δ, c=0.4δ, c=0.6p, c=0.2p, c=0.4p, c=0.6


Diplomatic Suspensions in Europe, 1820-1915

Suspension duration Frequency (%)(year)

1 350 (41.8)2 160 (19.1)3 61 (7.3)4 55 (6.6)5 33 (3.9)6 19 (2.3)7 15 (1.8)8 14 (1.7)9 9 (1.1)10 13 (1.6)> 10 102 (12.2)Total 837 (100)

Average 5.36 yearsMedian 2 years

1 Book Project


3 The Model

4 Equilibria



7 Conclusion

8 APPENDIX

Risk of War: Informational Effect

Pacifying (deterrent) effect

Diplomatic communication, when informative, always reduces theprob that S1 attacks:

1 − G (λ∗) ≥ 1 − G (κ∗d) > 1 − G (κ∗s ).

Resist

1

Surrender

22

m {r, s}

pv1 – c1(1 – p)v2 – c2

(v1, 0)(0, v2)Status quo

Attack

Risk of War: Commitment Effect

War-risking (commitment) effect

Diplomatic communication generally increases the probability thatS2 resists if attacked:

1 − F (β∗) > 1 − F (γ∗).

Resist

1

Surrender

22

m {r, s}

pv1 – c1(1 – p)v2 – c2


Attack

Risk of War: Net Effect

Diplomatic communication is designed to deter aggression,but its process creates a new commitment for State 2.

Deterrent effect vs. Commitment effect.

No analytical account of how these countervailing effectsbalance out to affect the overall probability of armed conflict.

Whether the deterrent effect dominates the commitmenteffect depends on parameter conditions.

→ Nonlinearity

Risk of War: Numerical Approach

Parameters Cutoff points Risk of war EquilibriumN δ c1 c2 p κ α β ωs,d ωb2 .9 .1 .1 .518 .193 .139 .139 .695 .691 Sincere2 .9 .1 .1 .800 .069 .110 .459 .503 .472 Deception5 .6 .1 .1 .364 .275 .099 .099 .654 .690 Sincere5 .6 .1 .1 .890 .053 .406 .704 .281 .090 Deception1 .9 .4 .4 .350 .211 .010 .611 .307 .306 Deception1 .9 .4 .4 .440 .910 .037 .037 .087 .247 Sincere1 .9 .4 .4 .500 .145 .165 .744 .219 .182 Deception

2 .8 .1 .2 .344 .291 .187 .187 .576 .606 Sincere2 .8 .1 .2 .750 .122 .493 .525 .417 .558 Deception

Risk of war is red if ωs,d < ωb or

Pr(war |informative) < Pr(war |babbling)

Risk of War: Numerical Approach

0.36

0.44 0.35 0.34

wb

0.89 0.35 0.59

Example: N=2, = 0.9, c1 = c2 = 0.4

wb

wb

Example: N=5, = 0.6, c1= c2 = 0.1

Example: N=2, = 0.8, c1 = 0.1, c2 = 0.2

0.60

wb

Example: N=1, = 0.9, c1 = c2 = 0.4

0.80

Functional form 4 Functional form 3

Functional form 1 Functional form 2

0

0 0

01

1 1

1

Four functional forms linking diplomacy, power, and war

Risk of War: Semiparametric Approach

Why empirics?

Use statistical analysis to supplement the theoreticalambiguity.

No unique prediction available (theoretically)

Not a test of a point prediction, rather an exploration of thedata to refine the prediction

To single out a unique ceteris paribus structure in the data, Iuse GAM to let the data “speak” for themselves about thefunctional form.

Research Design

What Exactly Are We Looking For?

Not the probability of war w/ and w/out diplomacy.

The impact of credible diplomatic communication on war.

Communication is credible in the informative equilibrium.

Research Design

Empirical Challenge

Measuring credible diplomatic precrisis communication

Hard to know whether precrisis communication took place for eachcrisis case.

Important sources of information are social networks with localofficials in the host governmentDinner parties, receptions, balls very crucialNewspapers or news wires do not pick up any of this

Research Design

Empirical Challenge

Measuring credible diplomatic precrisis communication

Hard to know if communication was credible.

Hard to know if a message was received and taken seriouslyHard to know if any given observation (of MIDs) belongs tothe informative equilibria, not the babbling equilibria.

Research Design

Empirical Solution

Use resident diplomatic missions as a proxy measure

The device for reliable diplomatic communication

Limit the empirical domain to European powers (plus the OttomanEmpire and U.S.)

1815-1914: Before communication revolutions (telegrams, jetliners)

Anecdotal evidence suggests this proxy is good until WWI (orperhaps today)

Model Specification

Dependent variable(s): Use of force in MIDs

coded as a “1” if both sides, A and B, use forcecoded as a “0” otherwise

Independent Variables

Diplomatic representation B → A before MIDResides in the capital at the rank of Chargé d’Affaires orabove.If the position is vacant, absent, located in other states, orconsulate coded as a “0”Interacted with p, balance of power

Controls

Diplomatic representation A → B before MIDInteracted with pOther factors that influence diplomatic exchange

Risk of War: Semiparametric Specification

Force = f1(cap. ratio× A → B) + f2(cap. ratio × B → A)+ f3(cap. ratio) + β4A → B + β5B → A+ Σki=6βixi + a + ε,

fi (·): Smooth function to estimateβi (·): Linear coefficient to estimateB → A: Diplomatic representation B to A (dummy)A → B: Diplomatic representation A to B (dummy)Σki=6xi : Set of controls

Risk of War: Semiparametric Approach

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Capabilities Ratio

Pro

babi

lity

of A

rmed

Con

flict

Diplomatic Mission A to B

0.0 0.2 0.4 0.6 0.8 1.00.

00.

20.

40.

60.

81.

0

Capabilities Ratio

Diplomatic Mission B to A

1 Book Project


3 The Model

4 Equilibria



7 Conclusion

8 APPENDIX

Conclusion

Precrisis communication is much more informative than previouslysuggested.

Defender’s credible precrisis communication reduces the risk of warroughly by 20% unless capabilities ratio is excessively skewed

Overall pacifying effect does not reject the war-risking effect ofthe commitment trap

Model and data here do not answer two questions:

Causal mechanism — Chapter 6Diplomacy increases the risk of war if a militarily weakchallenger uses it

Conclusion . . .

Big Picture

Diplomacy is not just the message that is sent, or the act of sendinga message

. . . rather an equilibrium in which such communication gainscredibility due to

Expectations of repeated diplomatic interactionExpectations of diplomatic break down

Diplomacy is more effective as an instrument of statecraft thanpreviously believed.

Thank you!

Policy Implications

Barack Obama’s commitment to diplomacy“It is important . . . to talk to . . . enemies. In fact, that’swhere diplomacy makes the biggest differences.”

Presidential debateClinton: “I believe in coercive diplomacy”McCain: “This is dangerous. It isn’t just naive; it’s dangerous”

The Model: Information, Beliefs, and History

Two-sided incomplete information.

Both player’s valuations are drawn from F (·) and G (·) eachperiod.

S1 updates its belief about v2, upon hearing S2’s diplomatictalk about its intention.

S1’s belief-updating at t is contingent on a history ht

Two kinds of history are relevant for history-dependent beliefs:

1 A history htC that induces the “normal” phase at t, where S2’sfailure to honor her commitment has not been detected in theprevious N periods.

2 A history htDN that invokes the “punishment” phase, whereS2’s failed commitment has been detected at t − N.

Babbling Equilibrium

Proposition 1 (Cheap-talk diplomacy)

There always exists a perfect Bayesian equilibrium, where in everyperiod each type of S2 announces anything or randomizes between“m = r” and “m = s,” and S1 attacks if v1 ≥ λ∗ or maintains thestatus quo if v1 < λ

∗, where λ∗ = [1−F (γ)]c1p+(1−p)F (γ) .

Sincere Equilibrium

Proposition 2 (Sincere equilibrium)

There exists a set of fully separating equilibria in stationarystrategies, in which:

all types with v2 ≥ α make a threat (m∗ = r) and resists ifattacked

all types with v2 < α make no threat (m∗ = s) and surrenders

if attacked

as long as S2 has an honesty reputation. Given the dishonestyreputation, players play the babbling equilibrium

Sincere Equilibrium

Sincere Equilibrium

If δ ≥ δ, the following strategies and beliefs constitute a perfect Bayesianequilibrium in trigger strategies. Providing hC , S2 announces that it willresist (i .e.,m = r) and resists when attacked if v2 ≥ α∗, and announcesm = s and surrenders when attacked if v2 < α

∗. Upon receiving m = r ,S1 believes that S2 always resists and attacks if v1 ≥ κ∗(r), while uponreceiving m = s, S1 believes that S2 never resists and hence attacksalways. Providing hDN for N = {1, 2, . . . ,∞}, both states play thebabbling equilibrium.

Deception Equilibrium

Proposition 3 (Deception equilibrium)

There exists a perfect Bayesian equilibrium in stationary strategiesin which S2 types are partitioned into three subsets provided thatS2 has a reputation for honesty

v2 ≥ β make a threat and resist if attackedv2 ∈ [α, β] make a threat but surrender if attackedv2 < α make no threat and surrender

Given the dishonesty reputation, players play the babblingequilibrium



If δ < δ, the following strategies and beliefs constitute a perfect Bayesianequilibrium in trigger strategies. Given hC , S2 plays: all types withv2 ≥ β∗ announce that m = r and resist if attacked; types withv2 ∈ [α∗, β∗] say m = r but surrender if attacked; and types withv2 < α

∗ say m = s and surrender if attacked. Upon receiving a messager , S1 believes that S2 will resist with probability (1− F (β∗))/(1− F (α∗))and attacks if v1 ≥ κ∗(r); upon receiving s, the posterior belief is zeroand hence S2 always attacks, where κ

∗(r) = ψ c1p . Given hDN , both statesplay the babbling equilibrium.

Lemma 1: Non-existence of a “false-positive” equilibrium

Another type of the equilibrium diplomatic message is not possible:

Lemma 1: Another type of informative equilibria

There is no PBE in which S2 falsely says m = s but actually resists ifattacked.

Research Design: anecdotal evidence

Telecommunication would invalidate diplomatic representation as aproxy measure

Impact of telecommunication

Queen Victoria opposed to upgrading British legation at Rometo the status of an embassy in 1876 since telegraphy hadrendered the expensive embassy obsolescent.

Impact of telegraphy was limited before WWI.

In 1860 the British MOFA instructed its diplomatic missionsnot to use telegrams unnecessarily since its cost is prohibitivelyhigh

Embassies as primary communication tool until after WWI

Receiving declaration of war by cable from Austria in 1914,Serbian PM Pas̆ić suspected the telegraph was a practical joke.Only a few hours later when the air-raid of Belgrade began didhe discovered that that was not a joke.

Data Collection

For testing the prediction, data collection just completed.

For each dyad involving European powers (and OttomanEmpire), gathered data on existence and rank of diplomaticrepresentatives of both directions for each year in 1815-1913

Utilized the automated text analysis and 15 undergrads

Source: Almanach de Gotha

Augments existing data sets with five-year interval by Singerand Small (1966) and Reşat Bayer (2006)

Independent Variable

Independent variable: Existence of diplomatic communication

coded as a “1” if

Chargé d’affairesMinister or EnvoyAmbassador

coded as a “0” if

ConsulsOther titlesNo evidence of diplomatic exchange

coded as a “-9” if

MissingOther titles

Model GAM GAM ProbitIndependent variable Resident mission Diplo. relations Resident mission

s(A → B× cap. ratio) 1.500(7.181)∗∗ 1.500(7.107)∗∗s(B → A× cap. ratio) 1.500(4.020)∗ 1.500(3.981)∗s(cap. ratio) 4.186 (0.899) 4.360 (0.828)cap. ratio −1.129(0.324)∗∗A → B 0.058 (0.041) 0.074 (0.041) 1.473 (0.484)B → A −0.092(0.034)∗∗ −0.095(0.034)∗∗ −0.522(0.197)∗∗Alliance 0.126 (0.096) 0.127 (0.097) 0.396 (0.287)Major power A -0.050 (0.071) -0.056 (0.071) -0.230 (0.197)Major power B −0.170(0.072)∗ −0.173(0.073)∗ −0.600(0.224)∗∗Joint democracy -0.010 (0.133) -0.011 (0.133) -0.115 (0.447)ln(peace years) −0.054(0.017)∗∗ −0.053(0.017)∗∗ −0.154(0.052)∗∗ln(distance) -0.022 (0.028) -0.021 (0.028) -0.093 (0.087)Contiguity 0.011 (0.068) 0.012 (0.068) 0.012 (0.203)Intercept 0.713(0.237)∗∗ 0.684(0.237)∗∗ 0.763 (0.838)Adjusted R2 0.123 0.123 0.117Deviance explained 16.2% 16.3% −193.8†GCV score 0.205 0.205 47.79‡

Probit estimates of diplomacy and war

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Capabilities Ratio

Pro

babi

lity

of F

orce

Diplomatic Mission A to B

0.0 0.2 0.4 0.6 0.8 1.00.

00.

20.

40.

60.

81.

0

Capabilities Ratio

Diplomatic Mission B to A

Book ProjectInformational Origins of Diplomacy and WarThe ModelEquilibriaCredibility of Diplomatic CommunicationDiplomacy and the Risk of WarConclusionAPPENDIX

diplomacy, precrisis communication, and war · diplomatic communication generally increases the...

Documents